Missing files in french update package
baptxste opened this issue · 6 comments
I'm trying to finetune the french model on personnal data but it seems that all files about rnnlm are missing :
/rnnlm
/exp/rnnlm
/exp/rnnlm_out
...
Is it possible to use the one provided in the en-model ? If not, do you have the files for the french model ?
If not, do you have the files for the french model ?
What model do you need an update package for? What model are you trying to modify?
I'm trying to modify this model https://alphacephei.com/vosk/models/vosk-model-fr-0.6-linto-2.2.0-compile.zip provided by the vosk website but the files mentioned above are missing.
I downloaded the english version(https://alphacephei.com/vosk/models/vosk-model-en-us-0.22-compile.zip) to compare and the files are present in this package. Can I use theses files in the french model to improve it ?
I'm trying to modify this model https://alphacephei.com/vosk/models/vosk-model-fr-0.6-linto-2.2.0-compile.zip provided by the vosk website but the files mentioned above are missing.
Linto model doesn't use llm rescoring, our vosk-model-fr does use it
I downloaded the english version(https://alphacephei.com/vosk/models/vosk-model-en-us-0.22-compile.zip) to compare and the files are present in this package. Can I use theses files in the french model to improve it ?
You need a model with rnnlm support, english files are not going to work in french model
Ok thanks for the clarification. The big french model (https://alphacephei.com/vosk/models/vosk-model-fr-0.22.zip ) seems to be rnnlm based but there is no compile package for it. Do you have one or can I use the script from the en-us compile package ?
Is it possible to finetune the acoustic model of the Linto compile package "easily" or do I need to finetune directly from kaldi ?
Do you have one or can I use the script from the en-us compile package ?
You can get it by writing us an email
Is it possible to finetune the acoustic model of the Linto compile package "easily" or do I need to finetune directly from kaldi
There is no easy finetuning for the acoustic model
Ok thanks I willl write you an email right now.